Support of Speech Class Tokens?

[Alok Gupta]

Does Dialogflow CX detect intent streaming API supports Speech class tokens?

For example, $OOV_CLASS_ALPHANUMERIC_SEQUENCE 

[Svetlana Pastukhova]

The $OOV_CLASS_ALPHANUMERIC_SEQUENCE is implicitly supported by Dialogflow CX's auto speech adaptation (when enabled), depending on the language. See the second bullet point in the Limitations section.

If you are planning to use regexp entities with voice inputs, make sure you follow all the requirements from the Regexp entities section on that page.

If your question is related to something else, please add more details about your use case. Thanks.

[Alok Gupta]

Thanks for your quick response! 

Yeah, we are planning to use regexp entities with voice inputs. Based on the requirements from the Regexp entities section, do you think it will be able to support capturing alphanumeric strings? We know gathering email over voice would be hard therefore planning to capture it in multiple steps:

1. Capture Email prefix

2. Capture Domain

3. Capture Domain suffix

When the user says the email prefix over voice it may have spaces in between. Is there a way to allow spaces in the Regex entity?

--

		ALOK GUPTA Architect, Emerging Solutions M: (562) 276-8980

--
© 2022 Google Inc. 1600 Amphitheatre Parkway, Mountain View, CA 94043
 
Email preferences: You received this email because you signed up for the Dialogflow CX Edition Discussion Google Group (dialogflow-cx...@googlegroups.com) to participate in discussions with other members of the Dialogflow community and the Dialogflow Team.
---
You received this message because you are subscribed to the Google Groups "Dialogflow CX Edition users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dialogflow-cx-editi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dialogflow-cx-edition-users/9e038a53-e397-48e4-bb78-86b0fd9ce716n%40googlegroups.com.

[Svetlana Pastukhova]

Yes, capturing alphanumerics with regexp entities over voice should work if you follow all the requirement. 

For spaces, you can add two entity entries: one matching a string with no spaces and another that allow for spaces. This is described in the Example section.

If you need to remove spaces, you can use the SUBSTITUTE system function or do it in your backend.

If you experience issues, please open a support case and provide the following:

- exported agent

- a complete scenario to reproduce the issue

- audio recordings with different alphanumeric sequences and corresponding expected and actual transcripts for each

- integration you use

- InputAudioConfig settings

- project ID and agent ID (or the URL to the agent)